What problem does it solve?
It prevents repeating past decisions and missing project conventions by forcing you to research existing memory, apply it during work, and persist any new learnings.
Core Features & Use Cases
- Memory-First Protocol: Enforces Research → Response/Execution → Learning for every task to keep outputs consistent with historical context.
- Deterministic Context Retrieval: Uses memory_search and memory_get to pull relevant prior notes before writing or answering about past work.
- Persistent Team Knowledge: Saves newly discovered conventions, architectural decisions, and user preferences back into the memory/ markdown store for future tasks.
Use Case: When implementing a new feature, you can reuse earlier architectural decisions and coding conventions already stored in MEMORY.md and memory/*.md, then automatically store any updated rules after the change.
Quick Start
Use the memory-manager skill to start any task by researching MEMORY.md and memory/*.md, then apply the retrieved conventions throughout the work and save new learnings at the end.